On the first day of the I / O 2022 developer conference, Google grandly launched a fully managed PostgreSQL database called alloydb Compared with Aurora PostgreSQL competitor of Amazon cloud service (AWS), Google claims that alloydb is twice as efficient In addition, under the same workload, the operation efficiency of alloydb can reach four times that of standard PostgreSQL, and the speed of analysis and query is also 100 times faster.
Developers familiar with the Google cloud ecosystem may not be unfamiliar with the fully managed PostgreSQL database service.
Previously, the company has provided cloudsql for PostgreSQL and spanner, and Google cloud's fully managed relational database service also provides a PostgreSQL compatible interface.
The core of alloydb is still based on the standard PostgreSQL database. Only in order to give full play to the strength of Google's own infrastructure, the development team modified its kernel and tried to keep it in the latest version.
After working at AWS for a long time, andI gutmans moved to Google in 2020 and served as director of database products and vice president of engineering.
He said that although Google has provided great help in helping enterprises migrate MySQL and PostgreSQL to the cloud, it still failed to fully take care of some customers who want to migrate their legacy databases (such as Oracle) to open source services.
The reason is that many enterprises use more than one cloud service provider and want to run anywhere as flexibly as possible. After years of delay, more and more customers are willing to invest resources to get rid of relevant constraints.
With the rise of Postgres (and the decline of MySQL) and becoming the de facto standard of open source relational databases, Google has become more motivated to promote customers to migrate to dedicated high-performance PostgreSQL services.
Gutmans added: many Google customers are looking to use their relational databases for analyzing use cases, so the alloydb development team has spent a lot of energy to ensure that Postgres can bring a better performance experience to these users.
During his work at AWS, he led the management of many AWS analysis services, had the opportunity to understand the importance and criticality of data to customers, and accumulated a deep technical background.
However, with the changing trend of the industry, it is no longer necessary to talk to front-line developers - even many customers come from business departments or as analysts.
While seeing the integration of the real world, he also felt that users wanted to get real-time insight from their data.
Back to the bottom of the technology, we can see that alloydb is built on Google's existing infrastructure and can peel computing and storage away - similar to the infrastructure layer running spanner, bigquery and almost all Google services.
In addition to alloydb, which focuses on PostgreSQL, related services have also occupied a considerable advantage in the competition. However, when trying to support multiple database engines / query languages, you don't always optimize everything.
Given that enterprises require postgre to migrate these legacy data, Google finally decided to be the top in this field. Through kernel level changes, the team has achieved linear expansion of more than 64 virtual cores.
In terms of analysis, alloydb team has also created a set of customized cache services based on machine learning to learn customers' access patterns. Then, the row format of Postgres is converted to the column format in memory to significantly improve the efficiency of execution.